Fast and Certified Bounding of Security-Constrained DCOPF via Interval Bound Propagation

Eren Tekeler; Huan Zhang; Samuel Chevalier; Xiangru Zhong

arxiv: 2511.15624 · v2 · submitted 2025-11-19 · 📡 eess.SY · cs.SY

Fast and Certified Bounding of Security-Constrained DCOPF via Interval Bound Propagation

Eren Tekeler , Xiangru Zhong , Huan Zhang , Samuel Chevalier This is my paper

Pith reviewed 2026-05-17 20:30 UTC · model grok-4.3

classification 📡 eess.SY cs.SY

keywords Security-Constrained DCOPFInterval Bound PropagationOptimal Power FlowCertified BoundsPower System OptimizationN-1 ContingenciesComputational Graph

0 comments

The pith

Representing SC-DCOPF as a computational graph lets interval bound propagation deliver certified bounds on optimal dispatch costs for large systems with many contingencies.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper models the security-constrained DC optimal power flow problem as a computational graph. It applies interval bound propagation to compute certified upper and lower bounds on the optimal objective value across the complete set of N-1 contingencies. This sidesteps the rapid slowdown of commercial solvers as both grid size and contingency count increase into the thousands. A sympathetic reader would care because the method supplies guaranteed ranges on minimum costs for secure and economic operation without solving every scenario to optimality.

Core claim

By casting the SC-DCOPF market-clearing problem as a computational graph, interval bound propagation computes certified bounds on the optimal objective value over the full set of N-1 contingencies. On small cases the mean gap to the true optimum stays below 3.98 percent, and the method scales efficiently to an 8,316-bus system with thousands of contingencies.

What carries the argument

Interval Bound Propagation applied to the computational graph representation of SC-DCOPF, which propagates interval bounds forward through the network constraints and objective function to certify the range containing the optimal value.

If this is right

Transmission operators obtain certified cost ranges for dispatch without optimizing each contingency individually.
The approach maintains useful bound tightness while scaling to real-world grid sizes and contingency counts.
Certified bounds can screen severe contingencies or supply initial information for exact solvers.
Market-clearing times remain practical even when the number of security constraints grows large.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The graph-plus-IBP construction could be reused for related problems such as stochastic or multi-period DCOPF if they admit similar representations.
Tighter bounds might result from combining IBP with other propagation techniques or warm-starting from prior solutions.
Similar certified bounding could be tested on AC power flow models after suitable linear or convex relaxations.

Load-bearing premise

The SC-DCOPF market-clearing problem can be represented as a computational graph to which Interval Bound Propagation applies directly and produces useful certified bounds on the optimal objective.

What would settle it

Solving an 8,316-bus case with thousands of contingencies exactly via a commercial solver and checking whether the gap between the IBP bounds and the true optimum exceeds 4 percent or whether total runtime is no longer substantially lower than full optimization.

Figures

Figures reproduced from arXiv: 2511.15624 by Eren Tekeler, Huan Zhang, Samuel Chevalier, Xiangru Zhong.

**Figure 3.** Figure 3: Distribution of IBP mean runtimes in log scale across different system [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

read the original abstract

Security-Constrained DC Optimal Power Flow (SC DCOPF) is an important tool for transmission system operators, enabling economically efficient and physically secure dispatch decisions. Although CPU-based commercial solvers (e.g., Gurobi) can efficiently solve SC-DCOPF problems with a reasonable number of security constraints, their performance degrades rapidly as both system size and the number of contingencies grow into thousands. In this paper, we design a computational graph representation of the SC-DCOPF-based market-clearing problem, inspired by the third ARPA-E Grid Optimization Competition. Using a tool from the neural network verification community known as Interval Bound Propagation (IBP), we quickly compute bounds on the optimal objective across the full set of N-1 contingencies. Our results demonstrate that IBP can compute certified bounds with mean optimal solution gaps below 3.98% on small cases, and it can efficiently scale up to 8,316 bus systems with thousands of contingencies.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

IBP from neural net verification gives fast certified bounds on large SC-DCOPF instances, but the paper needs tighter validation against exact solvers and clearer graph construction details.

read the letter

The main point is that this paper takes interval bound propagation, a method developed for verifying neural networks, and applies it to compute certified bounds on the optimal objective of security-constrained DC optimal power flow problems. They encode the market-clearing formulation as a computational graph and propagate bounds across N-1 contingencies, reporting mean gaps under 3.98 percent on small cases while scaling to an 8,316-bus system with thousands of contingencies. That scaling claim is the part worth paying attention to, since standard solvers slow down sharply as contingency counts rise. The cross-domain move is genuinely new; IBP has not been a standard tool in power-systems optimization literature for this task. The work does a clean job of showing that the approach runs efficiently on realistically sized grids and produces usable certified bounds rather than just heuristic estimates. The computational-graph framing also aligns with the ARPA-E competition style, which helps reproducibility. The soft spots are mostly around validation depth. The abstract states the gap numbers but does not yet show detailed comparisons against exact solvers or other bounding methods on the larger instances, so it is not clear how much tightness is lost when the system grows. If the graph encoding omits certain constraint interactions or if bound looseness varies strongly with contingency selection, the practical value shrinks. Minor gaps in error analysis and baseline tables are fixable but should be addressed before final publication. This paper is aimed at power-systems researchers who need fast, certified bounds for real-time or market applications, and at optimization people interested in verification techniques. A reader working on large-scale SC-DCOPF or on transferring bound-propagation methods would find concrete scaling evidence here. The work deserves a serious referee because the central claim is testable, the method is formally grounded, and the scaling results are specific enough to evaluate. I would send it to review with requests for expanded comparisons and explicit checks on graph fidelity.

Referee Report

2 major / 3 minor

Summary. The manuscript proposes encoding the Security-Constrained DC Optimal Power Flow (SC-DCOPF) market-clearing problem as a computational graph and applying Interval Bound Propagation (IBP) to obtain certified bounds on the optimal objective value over the full set of N-1 contingencies. It reports that the resulting bounds exhibit mean optimality gaps below 3.98% on small test cases and demonstrates computational scalability on an 8,316-bus system with thousands of contingencies.

Significance. If the graph encoding is faithful and the reported gaps are reproducible, the work offers a promising cross-disciplinary technique for rapid, certified bounding of large-scale SC-DCOPF instances without invoking a full commercial solver for every contingency. The scalability result on an 8k-bus network is a concrete strength that could be of practical interest to transmission operators.

major comments (2)

[Section 3] The central claim that IBP produces certified bounds on the SC-DCOPF optimum rests on the assumption that the market-clearing problem can be represented exactly as a computational graph to which standard IBP applies without additional relaxations. The manuscript should explicitly state, in the section describing the graph construction, whether any linearization, big-M reformulation, or other approximation is introduced and how it affects the certified character of the bounds.
[Results section / Table 2] Table 2 (or the equivalent results table for the small cases) reports a mean gap below 3.98% but does not indicate the number of instances, the distribution of gaps, or the solver used to obtain the reference optima. Without these details the tightness claim cannot be fully evaluated.

minor comments (3)

[Abstract] The abstract cites inspiration from the third ARPA-E Grid Optimization Competition but provides no reference; please add the appropriate citation.
[Figures] Figure captions should explicitly state the system size, number of contingencies, and whether the plotted bounds are lower/upper or both.
[Section 2] Notation for the interval bounds (e.g., the symbols used for lower and upper bounds on the objective) should be introduced once in the methods and used consistently thereafter.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful review and positive recommendation for minor revision. We address each of the major comments below and have incorporated revisions to enhance the clarity of our manuscript.

read point-by-point responses

Referee: [Section 3] The central claim that IBP produces certified bounds on the SC-DCOPF optimum rests on the assumption that the market-clearing problem can be represented exactly as a computational graph to which standard IBP applies without additional relaxations. The manuscript should explicitly state, in the section describing the graph construction, whether any linearization, big-M reformulation, or other approximation is introduced and how it affects the certified character of the bounds.

Authors: We appreciate the referee's emphasis on this foundational aspect. The computational graph in Section 3 is constructed as an exact representation of the SC-DCOPF linear program, consisting of affine transformations for the objective and constraints corresponding to the DC power flow equations and contingency constraints, without any linearizations, big-M reformulations, or additional approximations. Standard IBP is applied directly to this graph, yielding certified bounds on the optimal objective value of the encoded problem. We have revised Section 3 to include an explicit statement confirming this exact encoding and explaining that the certified bounds hold with respect to the modeled SC-DCOPF instance. revision: yes
Referee: [Results section / Table 2] Table 2 (or the equivalent results table for the small cases) reports a mean gap below 3.98% but does not indicate the number of instances, the distribution of gaps, or the solver used to obtain the reference optima. Without these details the tightness claim cannot be fully evaluated.

Authors: We agree that providing these details will improve the transparency of our results. In the revised version, we have updated the results section and the caption of Table 2 to report the number of small test instances evaluated, the full distribution of optimality gaps (including range and standard deviation), and that the reference optimal costs were obtained using the Gurobi optimizer. These additions allow readers to better assess the tightness of the IBP bounds. revision: yes

Circularity Check

0 steps flagged

No significant circularity: direct application of existing IBP to SC-DCOPF graph

full rationale

The paper encodes the SC-DCOPF market-clearing problem as a computational graph and applies the pre-existing Interval Bound Propagation method from the neural-network verification literature to compute certified bounds on the optimal objective. No step in the provided abstract or description reduces a claimed result to a fitted parameter, self-defined quantity, or load-bearing self-citation; the tightness is instead validated by direct numerical comparison on small instances and scalability is shown on an 8316-bus system. The derivation therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review is based solely on the abstract; therefore the ledger is necessarily incomplete. The central claim rests on the standard DC power-flow approximation and the assumption that the market-clearing problem admits a computational-graph representation suitable for bound propagation.

axioms (1)

domain assumption DC power flow approximations are sufficiently accurate for the SC-DCOPF instances considered.
This is the modeling foundation of any DCOPF formulation and is implicit in the abstract.

pith-pipeline@v0.9.0 · 5470 in / 1235 out tokens · 72788 ms · 2026-05-17T20:30:38.893336+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

8 extracted references · 8 canonical work pages

[1]

J. T. Holzer, S. Elbert, H. Mittelmann, R. O’Neill, and H. Oh,Go competition challenge 3: Problem, solvers, and solution analysis, 2024. arXiv: 2411.12033[math.NA]. [Online]. Available: https://arxiv.org/ abs/2411.12033

work page arXiv 2024
[2]

A parallelized, adam-based solver for reserve and security constrained ac unit commitment,

S. Chevalier, “A parallelized, adam-based solver for reserve and security constrained ac unit commitment,”Electric Power Systems Research, vol. 235, p. 110 685, 2024,ISSN: 0378-7796.DOI: https://doi.org/10. 1016/j.epsr.2024.110685 [Online]. Available: https://www.sciencedirect. com/science/article/pii/S0378779624005716

work page arXiv 2024
[3]

On the effectiveness of interval bound propagation for training verifiably robust models

S. Gowal et al.,On the effectiveness of interval bound propagation for training verifiably robust models, 2019. arXiv: 1810.12715[cs.LG]. [Online]. Available: https://arxiv.org/abs/1810.12715

work page arXiv 2019
[4]

Ef- ficient neural network robustness certification with general activation functions,

H. Zhang, T.-W. Weng, P.-Y . Chen, C.-J. Hsieh, and L. Daniel, “Ef- ficient neural network robustness certification with general activation functions,”Advances in Neural Information Processing Systems, vol. 31, pp. 4939–4948, 2018. [Online]. Available: https://arxiv.org/pdf/1811. 00866.pdf

work page 2018
[5]

Fast and Complete: Enabling complete neural network verification with rapid and massively parallel incomplete verifiers,

K. Xu et al., “Fast and Complete: Enabling complete neural network verification with rapid and massively parallel incomplete verifiers,” in International Conference on Learning Representations, 2021. [Online]. Available: https://openreview.net/forum?id=nVZtXBI6LNn

work page 2021
[6]

A branch and bound framework for stronger adversarial attacks of ReLU networks,

H. Zhang et al., “A branch and bound framework for stronger adversarial attacks of ReLU networks,” inProceedings of the 39th International Conference on Machine Learning, vol. 162, 2022, pp. 26 591–26 604

work page 2022
[7]

Automatic perturbation analysis for scalable certified robustness and beyond,

K. Xu et al., “Automatic perturbation analysis for scalable certified robustness and beyond,”Advances in Neural Information Processing Systems, vol. 33, 2020

work page 2020
[8]

Horn and C

R. Horn and C. Johnson,Matrix Analysis. Cambridge University Press, 1990,ISBN: 9780521386326

work page 1990

[1] [1]

J. T. Holzer, S. Elbert, H. Mittelmann, R. O’Neill, and H. Oh,Go competition challenge 3: Problem, solvers, and solution analysis, 2024. arXiv: 2411.12033[math.NA]. [Online]. Available: https://arxiv.org/ abs/2411.12033

work page arXiv 2024

[2] [2]

A parallelized, adam-based solver for reserve and security constrained ac unit commitment,

S. Chevalier, “A parallelized, adam-based solver for reserve and security constrained ac unit commitment,”Electric Power Systems Research, vol. 235, p. 110 685, 2024,ISSN: 0378-7796.DOI: https://doi.org/10. 1016/j.epsr.2024.110685 [Online]. Available: https://www.sciencedirect. com/science/article/pii/S0378779624005716

work page arXiv 2024

[3] [3]

On the effectiveness of interval bound propagation for training verifiably robust models

S. Gowal et al.,On the effectiveness of interval bound propagation for training verifiably robust models, 2019. arXiv: 1810.12715[cs.LG]. [Online]. Available: https://arxiv.org/abs/1810.12715

work page arXiv 2019

[4] [4]

Ef- ficient neural network robustness certification with general activation functions,

H. Zhang, T.-W. Weng, P.-Y . Chen, C.-J. Hsieh, and L. Daniel, “Ef- ficient neural network robustness certification with general activation functions,”Advances in Neural Information Processing Systems, vol. 31, pp. 4939–4948, 2018. [Online]. Available: https://arxiv.org/pdf/1811. 00866.pdf

work page 2018

[5] [5]

Fast and Complete: Enabling complete neural network verification with rapid and massively parallel incomplete verifiers,

K. Xu et al., “Fast and Complete: Enabling complete neural network verification with rapid and massively parallel incomplete verifiers,” in International Conference on Learning Representations, 2021. [Online]. Available: https://openreview.net/forum?id=nVZtXBI6LNn

work page 2021

[6] [6]

A branch and bound framework for stronger adversarial attacks of ReLU networks,

H. Zhang et al., “A branch and bound framework for stronger adversarial attacks of ReLU networks,” inProceedings of the 39th International Conference on Machine Learning, vol. 162, 2022, pp. 26 591–26 604

work page 2022

[7] [7]

Automatic perturbation analysis for scalable certified robustness and beyond,

K. Xu et al., “Automatic perturbation analysis for scalable certified robustness and beyond,”Advances in Neural Information Processing Systems, vol. 33, 2020

work page 2020

[8] [8]

Horn and C

R. Horn and C. Johnson,Matrix Analysis. Cambridge University Press, 1990,ISBN: 9780521386326

work page 1990